SAFE public projects git trees. - safe/jmp/linux-2.6/blob - net/ipv6/sit.c

   1 /*
   2  *      IPv6 over IPv4 tunnel device - Simple Internet Transition (SIT)
   3  *      Linux INET6 implementation
   4  *
   5  *      Authors:
   6  *      Pedro Roque             <roque@di.fc.ul.pt>
   7  *      Alexey Kuznetsov        <kuznet@ms2.inr.ac.ru>
   8  *
   9  *      $Id: sit.c,v 1.53 2001/09/25 05:09:53 davem Exp $
  10  *
  11  *      This program is free software; you can redistribute it and/or
  12  *      modify it under the terms of the GNU General Public License
  13  *      as published by the Free Software Foundation; either version
  14  *      2 of the License, or (at your option) any later version.
  15  *
  16  *      Changes:
  17  * Roger Venning <r.venning@telstra.com>:       6to4 support
  18  * Nate Thompson <nate@thebog.net>:             6to4 support
  19  * Fred Templin <fred.l.templin@boeing.com>:    isatap support
  20  */
  21
  22 #include <linux/module.h>
  23 #include <linux/capability.h>
  24 #include <linux/errno.h>
  25 #include <linux/types.h>
  26 #include <linux/socket.h>
  27 #include <linux/sockios.h>
  28 #include <linux/net.h>
  29 #include <linux/in6.h>
  30 #include <linux/netdevice.h>
  31 #include <linux/if_arp.h>
  32 #include <linux/icmp.h>
  33 #include <asm/uaccess.h>
  34 #include <linux/init.h>
  35 #include <linux/netfilter_ipv4.h>
  36 #include <linux/if_ether.h>
  37
  38 #include <net/sock.h>
  39 #include <net/snmp.h>
  40
  41 #include <net/ipv6.h>
  42 #include <net/protocol.h>
  43 #include <net/transp_v6.h>
  44 #include <net/ip6_fib.h>
  45 #include <net/ip6_route.h>
  46 #include <net/ndisc.h>
  47 #include <net/addrconf.h>
  48 #include <net/ip.h>
  49 #include <net/udp.h>
  50 #include <net/icmp.h>
  51 #include <net/ipip.h>
  52 #include <net/inet_ecn.h>
  53 #include <net/xfrm.h>
  54 #include <net/dsfield.h>
  55 #include <net/net_namespace.h>
  56 #include <net/netns/generic.h>
  57
  58 /*
  59    This version of net/ipv6/sit.c is cloned of net/ipv4/ip_gre.c
  60
  61    For comments look at net/ipv4/ip_gre.c --ANK
  62  */
  63
  64 #define HASH_SIZE  16
  65 #define HASH(addr) (((__force u32)addr^((__force u32)addr>>4))&0xF)
  66
  67 static int ipip6_fb_tunnel_init(struct net_device *dev);
  68 static int ipip6_tunnel_init(struct net_device *dev);
  69 static void ipip6_tunnel_setup(struct net_device *dev);
  70
  71 static int sit_net_id;
  72 struct sit_net {
  73 };
  74
  75 static struct net_device *ipip6_fb_tunnel_dev;
  76
  77 static struct ip_tunnel *tunnels_r_l[HASH_SIZE];
  78 static struct ip_tunnel *tunnels_r[HASH_SIZE];
  79 static struct ip_tunnel *tunnels_l[HASH_SIZE];
  80 static struct ip_tunnel *tunnels_wc[1];
  81 static struct ip_tunnel **tunnels[4] = { tunnels_wc, tunnels_l, tunnels_r, tunnels_r_l };
  82
  83 static DEFINE_RWLOCK(ipip6_lock);
  84
  85 static struct ip_tunnel * ipip6_tunnel_lookup(__be32 remote, __be32 local)
  86 {
  87         unsigned h0 = HASH(remote);
  88         unsigned h1 = HASH(local);
  89         struct ip_tunnel *t;
  90
  91         for (t = tunnels_r_l[h0^h1]; t; t = t->next) {
  92                 if (local == t->parms.iph.saddr &&
  93                     remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  94                         return t;
  95         }
  96         for (t = tunnels_r[h0]; t; t = t->next) {
  97                 if (remote == t->parms.iph.daddr && (t->dev->flags&IFF_UP))
  98                         return t;
  99         }
 100         for (t = tunnels_l[h1]; t; t = t->next) {
 101                 if (local == t->parms.iph.saddr && (t->dev->flags&IFF_UP))
 102                         return t;
 103         }
 104         if ((t = tunnels_wc[0]) != NULL && (t->dev->flags&IFF_UP))
 105                 return t;
 106         return NULL;
 107 }
 108
 109 static struct ip_tunnel **__ipip6_bucket(struct ip_tunnel_parm *parms)
 110 {
 111         __be32 remote = parms->iph.daddr;
 112         __be32 local = parms->iph.saddr;
 113         unsigned h = 0;
 114         int prio = 0;
 115
 116         if (remote) {
 117                 prio |= 2;
 118                 h ^= HASH(remote);
 119         }
 120         if (local) {
 121                 prio |= 1;
 122                 h ^= HASH(local);
 123         }
 124         return &tunnels[prio][h];
 125 }
 126
 127 static inline struct ip_tunnel **ipip6_bucket(struct ip_tunnel *t)
 128 {
 129         return __ipip6_bucket(&t->parms);
 130 }
 131
 132 static void ipip6_tunnel_unlink(struct ip_tunnel *t)
 133 {
 134         struct ip_tunnel **tp;
 135
 136         for (tp = ipip6_bucket(t); *tp; tp = &(*tp)->next) {
 137                 if (t == *tp) {
 138                         write_lock_bh(&ipip6_lock);
 139                         *tp = t->next;
 140                         write_unlock_bh(&ipip6_lock);
 141                         break;
 142                 }
 143         }
 144 }
 145
 146 static void ipip6_tunnel_link(struct ip_tunnel *t)
 147 {
 148         struct ip_tunnel **tp = ipip6_bucket(t);
 149
 150         t->next = *tp;
 151         write_lock_bh(&ipip6_lock);
 152         *tp = t;
 153         write_unlock_bh(&ipip6_lock);
 154 }
 155
 156 static struct ip_tunnel * ipip6_tunnel_locate(struct ip_tunnel_parm *parms, int create)
 157 {
 158         __be32 remote = parms->iph.daddr;
 159         __be32 local = parms->iph.saddr;
 160         struct ip_tunnel *t, **tp, *nt;
 161         struct net_device *dev;
 162         char name[IFNAMSIZ];
 163
 164         for (tp = __ipip6_bucket(parms); (t = *tp) != NULL; tp = &t->next) {
 165                 if (local == t->parms.iph.saddr && remote == t->parms.iph.daddr)
 166                         return t;
 167         }
 168         if (!create)
 169                 goto failed;
 170
 171         if (parms->name[0])
 172                 strlcpy(name, parms->name, IFNAMSIZ);
 173         else
 174                 sprintf(name, "sit%%d");
 175
 176         dev = alloc_netdev(sizeof(*t), name, ipip6_tunnel_setup);
 177         if (dev == NULL)
 178                 return NULL;
 179
 180         if (strchr(name, '%')) {
 181                 if (dev_alloc_name(dev, name) < 0)
 182                         goto failed_free;
 183         }
 184
 185         nt = netdev_priv(dev);
 186         dev->init = ipip6_tunnel_init;
 187         nt->parms = *parms;
 188
 189         if (parms->i_flags & SIT_ISATAP)
 190                 dev->priv_flags |= IFF_ISATAP;
 191
 192         if (register_netdevice(dev) < 0)
 193                 goto failed_free;
 194
 195         dev_hold(dev);
 196
 197         ipip6_tunnel_link(nt);
 198         return nt;
 199
 200 failed_free:
 201         free_netdev(dev);
 202 failed:
 203         return NULL;
 204 }
 205
 206 static struct ip_tunnel_prl_entry *
 207 __ipip6_tunnel_locate_prl(struct ip_tunnel *t, __be32 addr)
 208 {
 209         struct ip_tunnel_prl_entry *p = (struct ip_tunnel_prl_entry *)NULL;
 210
 211         for (p = t->prl; p; p = p->next)
 212                 if (p->addr == addr)
 213                         break;
 214         return p;
 215
 216 }
 217
 218 static int ipip6_tunnel_get_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 219 {
 220         struct ip_tunnel_prl *kp;
 221         struct ip_tunnel_prl_entry *prl;
 222         unsigned int cmax, c = 0, ca, len;
 223         int ret = 0;
 224
 225         cmax = a->datalen / sizeof(*a);
 226         if (cmax > 1 && a->addr != htonl(INADDR_ANY))
 227                 cmax = 1;
 228
 229         /* For simple GET or for root users,
 230          * we try harder to allocate.
 231          */
 232         kp = (cmax <= 1 || capable(CAP_NET_ADMIN)) ?
 233                 kcalloc(cmax, sizeof(*kp), GFP_KERNEL) :
 234                 NULL;
 235
 236         read_lock(&ipip6_lock);
 237
 238         ca = t->prl_count < cmax ? t->prl_count : cmax;
 239
 240         if (!kp) {
 241                 /* We don't try hard to allocate much memory for
 242                  * non-root users.
 243                  * For root users, retry allocating enough memory for
 244                  * the answer.
 245                  */
 246                 kp = kcalloc(ca, sizeof(*kp), GFP_ATOMIC);
 247                 if (!kp) {
 248                         ret = -ENOMEM;
 249                         goto out;
 250                 }
 251         }
 252
 253         c = 0;
 254         for (prl = t->prl; prl; prl = prl->next) {
 255                 if (c > cmax)
 256                         break;
 257                 if (a->addr != htonl(INADDR_ANY) && prl->addr != a->addr)
 258                         continue;
 259                 kp[c].addr = prl->addr;
 260                 kp[c].flags = prl->flags;
 261                 c++;
 262                 if (a->addr != htonl(INADDR_ANY))
 263                         break;
 264         }
 265 out:
 266         read_unlock(&ipip6_lock);
 267
 268         len = sizeof(*kp) * c;
 269         ret = len ? copy_to_user(a->data, kp, len) : 0;
 270
 271         kfree(kp);
 272         if (ret)
 273                 return -EFAULT;
 274
 275         a->datalen = len;
 276         return 0;
 277 }
 278
 279 static int
 280 ipip6_tunnel_add_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a, int chg)
 281 {
 282         struct ip_tunnel_prl_entry *p;
 283         int err = 0;
 284
 285         if (a->addr == htonl(INADDR_ANY))
 286                 return -EINVAL;
 287
 288         write_lock(&ipip6_lock);
 289
 290         for (p = t->prl; p; p = p->next) {
 291                 if (p->addr == a->addr) {
 292                         if (chg)
 293                                 goto update;
 294                         err = -EEXIST;
 295                         goto out;
 296                 }
 297         }
 298
 299         if (chg) {
 300                 err = -ENXIO;
 301                 goto out;
 302         }
 303
 304         p = kzalloc(sizeof(struct ip_tunnel_prl_entry), GFP_KERNEL);
 305         if (!p) {
 306                 err = -ENOBUFS;
 307                 goto out;
 308         }
 309
 310         p->next = t->prl;
 311         t->prl = p;
 312         t->prl_count++;
 313 update:
 314         p->addr = a->addr;
 315         p->flags = a->flags;
 316 out:
 317         write_unlock(&ipip6_lock);
 318         return err;
 319 }
 320
 321 static int
 322 ipip6_tunnel_del_prl(struct ip_tunnel *t, struct ip_tunnel_prl *a)
 323 {
 324         struct ip_tunnel_prl_entry *x, **p;
 325         int err = 0;
 326
 327         write_lock(&ipip6_lock);
 328
 329         if (a && a->addr != htonl(INADDR_ANY)) {
 330                 for (p = &t->prl; *p; p = &(*p)->next) {
 331                         if ((*p)->addr == a->addr) {
 332                                 x = *p;
 333                                 *p = x->next;
 334                                 kfree(x);
 335                                 t->prl_count--;
 336                                 goto out;
 337                         }
 338                 }
 339                 err = -ENXIO;
 340         } else {
 341                 while (t->prl) {
 342                         x = t->prl;
 343                         t->prl = t->prl->next;
 344                         kfree(x);
 345                         t->prl_count--;
 346                 }
 347         }
 348 out:
 349         write_unlock(&ipip6_lock);
 350         return 0;
 351 }
 352
 353 static int
 354 isatap_chksrc(struct sk_buff *skb, struct iphdr *iph, struct ip_tunnel *t)
 355 {
 356         struct ip_tunnel_prl_entry *p;
 357         int ok = 1;
 358
 359         read_lock(&ipip6_lock);
 360         p = __ipip6_tunnel_locate_prl(t, iph->saddr);
 361         if (p) {
 362                 if (p->flags & PRL_DEFAULT)
 363                         skb->ndisc_nodetype = NDISC_NODETYPE_DEFAULT;
 364                 else
 365                         skb->ndisc_nodetype = NDISC_NODETYPE_NODEFAULT;
 366         } else {
 367                 struct in6_addr *addr6 = &ipv6_hdr(skb)->saddr;
 368                 if (ipv6_addr_is_isatap(addr6) &&
 369                     (addr6->s6_addr32[3] == iph->saddr) &&
 370                     ipv6_chk_prefix(addr6, t->dev))
 371                         skb->ndisc_nodetype = NDISC_NODETYPE_HOST;
 372                 else
 373                         ok = 0;
 374         }
 375         read_unlock(&ipip6_lock);
 376         return ok;
 377 }
 378
 379 static void ipip6_tunnel_uninit(struct net_device *dev)
 380 {
 381         if (dev == ipip6_fb_tunnel_dev) {
 382                 write_lock_bh(&ipip6_lock);
 383                 tunnels_wc[0] = NULL;
 384                 write_unlock_bh(&ipip6_lock);
 385                 dev_put(dev);
 386         } else {
 387                 ipip6_tunnel_unlink(netdev_priv(dev));
 388                 ipip6_tunnel_del_prl(netdev_priv(dev), NULL);
 389                 dev_put(dev);
 390         }
 391 }
 392
 393
 394 static int ipip6_err(struct sk_buff *skb, u32 info)
 395 {
 396 #ifndef I_WISH_WORLD_WERE_PERFECT
 397
 398 /* It is not :-( All the routers (except for Linux) return only
 399    8 bytes of packet payload. It means, that precise relaying of
 400    ICMP in the real Internet is absolutely infeasible.
 401  */
 402         struct iphdr *iph = (struct iphdr*)skb->data;
 403         const int type = icmp_hdr(skb)->type;
 404         const int code = icmp_hdr(skb)->code;
 405         struct ip_tunnel *t;
 406         int err;
 407
 408         switch (type) {
 409         default:
 410         case ICMP_PARAMETERPROB:
 411                 return 0;
 412
 413         case ICMP_DEST_UNREACH:
 414                 switch (code) {
 415                 case ICMP_SR_FAILED:
 416                 case ICMP_PORT_UNREACH:
 417                         /* Impossible event. */
 418                         return 0;
 419                 case ICMP_FRAG_NEEDED:
 420                         /* Soft state for pmtu is maintained by IP core. */
 421                         return 0;
 422                 default:
 423                         /* All others are translated to HOST_UNREACH.
 424                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 425                            I believe they are just ether pollution. --ANK
 426                          */
 427                         break;
 428                 }
 429                 break;
 430         case ICMP_TIME_EXCEEDED:
 431                 if (code != ICMP_EXC_TTL)
 432                         return 0;
 433                 break;
 434         }
 435
 436         err = -ENOENT;
 437
 438         read_lock(&ipip6_lock);
 439         t = ipip6_tunnel_lookup(iph->daddr, iph->saddr);
 440         if (t == NULL || t->parms.iph.daddr == 0)
 441                 goto out;
 442
 443         err = 0;
 444         if (t->parms.iph.ttl == 0 && type == ICMP_TIME_EXCEEDED)
 445                 goto out;
 446
 447         if (jiffies - t->err_time < IPTUNNEL_ERR_TIMEO)
 448                 t->err_count++;
 449         else
 450                 t->err_count = 1;
 451         t->err_time = jiffies;
 452 out:
 453         read_unlock(&ipip6_lock);
 454         return err;
 455 #else
 456         struct iphdr *iph = (struct iphdr*)dp;
 457         int hlen = iph->ihl<<2;
 458         struct ipv6hdr *iph6;
 459         const int type = icmp_hdr(skb)->type;
 460         const int code = icmp_hdr(skb)->code;
 461         int rel_type = 0;
 462         int rel_code = 0;
 463         int rel_info = 0;
 464         struct sk_buff *skb2;
 465         struct rt6_info *rt6i;
 466
 467         if (len < hlen + sizeof(struct ipv6hdr))
 468                 return;
 469         iph6 = (struct ipv6hdr*)(dp + hlen);
 470
 471         switch (type) {
 472         default:
 473                 return;
 474         case ICMP_PARAMETERPROB:
 475                 if (icmp_hdr(skb)->un.gateway < hlen)
 476                         return;
 477
 478                 /* So... This guy found something strange INSIDE encapsulated
 479                    packet. Well, he is fool, but what can we do ?
 480                  */
 481                 rel_type = ICMPV6_PARAMPROB;
 482                 rel_info = icmp_hdr(skb)->un.gateway - hlen;
 483                 break;
 484
 485         case ICMP_DEST_UNREACH:
 486                 switch (code) {
 487                 case ICMP_SR_FAILED:
 488                 case ICMP_PORT_UNREACH:
 489                         /* Impossible event. */
 490                         return;
 491                 case ICMP_FRAG_NEEDED:
 492                         /* Too complicated case ... */
 493                         return;
 494                 default:
 495                         /* All others are translated to HOST_UNREACH.
 496                            rfc2003 contains "deep thoughts" about NET_UNREACH,
 497                            I believe, it is just ether pollution. --ANK
 498                          */
 499                         rel_type = ICMPV6_DEST_UNREACH;
 500                         rel_code = ICMPV6_ADDR_UNREACH;
 501                         break;
 502                 }
 503                 break;
 504         case ICMP_TIME_EXCEEDED:
 505                 if (code != ICMP_EXC_TTL)
 506                         return;
 507                 rel_type = ICMPV6_TIME_EXCEED;
 508                 rel_code = ICMPV6_EXC_HOPLIMIT;
 509                 break;
 510         }
 511
 512         /* Prepare fake skb to feed it to icmpv6_send */
 513         skb2 = skb_clone(skb, GFP_ATOMIC);
 514         if (skb2 == NULL)
 515                 return 0;
 516         dst_release(skb2->dst);
 517         skb2->dst = NULL;
 518         skb_pull(skb2, skb->data - (u8*)iph6);
 519         skb_reset_network_header(skb2);
 520
 521         /* Try to guess incoming interface */
 522         rt6i = rt6_lookup(&init_net, &iph6->saddr, NULL, NULL, 0);
 523         if (rt6i && rt6i->rt6i_dev) {
 524                 skb2->dev = rt6i->rt6i_dev;
 525
 526                 rt6i = rt6_lookup(&init_net, &iph6->daddr, &iph6->saddr, NULL, 0);
 527
 528                 if (rt6i && rt6i->rt6i_dev && rt6i->rt6i_dev->type == ARPHRD_SIT) {
 529                         struct ip_tunnel *t = netdev_priv(rt6i->rt6i_dev);
 530                         if (rel_type == ICMPV6_TIME_EXCEED && t->parms.iph.ttl) {
 531                                 rel_type = ICMPV6_DEST_UNREACH;
 532                                 rel_code = ICMPV6_ADDR_UNREACH;
 533                         }
 534                         icmpv6_send(skb2, rel_type, rel_code, rel_info, skb2->dev);
 535                 }
 536         }
 537         kfree_skb(skb2);
 538         return 0;
 539 #endif
 540 }
 541
 542 static inline void ipip6_ecn_decapsulate(struct iphdr *iph, struct sk_buff *skb)
 543 {
 544         if (INET_ECN_is_ce(iph->tos))
 545                 IP6_ECN_set_ce(ipv6_hdr(skb));
 546 }
 547
 548 static int ipip6_rcv(struct sk_buff *skb)
 549 {
 550         struct iphdr *iph;
 551         struct ip_tunnel *tunnel;
 552
 553         if (!pskb_may_pull(skb, sizeof(struct ipv6hdr)))
 554                 goto out;
 555
 556         iph = ip_hdr(skb);
 557
 558         read_lock(&ipip6_lock);
 559         if ((tunnel = ipip6_tunnel_lookup(iph->saddr, iph->daddr)) != NULL) {
 560                 secpath_reset(skb);
 561                 skb->mac_header = skb->network_header;
 562                 skb_reset_network_header(skb);
 563                 IPCB(skb)->flags = 0;
 564                 skb->protocol = htons(ETH_P_IPV6);
 565                 skb->pkt_type = PACKET_HOST;
 566
 567                 if ((tunnel->dev->priv_flags & IFF_ISATAP) &&
 568                     !isatap_chksrc(skb, iph, tunnel)) {
 569                         tunnel->stat.rx_errors++;
 570                         read_unlock(&ipip6_lock);
 571                         kfree_skb(skb);
 572                         return 0;
 573                 }
 574                 tunnel->stat.rx_packets++;
 575                 tunnel->stat.rx_bytes += skb->len;
 576                 skb->dev = tunnel->dev;
 577                 dst_release(skb->dst);
 578                 skb->dst = NULL;
 579                 nf_reset(skb);
 580                 ipip6_ecn_decapsulate(iph, skb);
 581                 netif_rx(skb);
 582                 read_unlock(&ipip6_lock);
 583                 return 0;
 584         }
 585
 586         icmp_send(skb, ICMP_DEST_UNREACH, ICMP_PORT_UNREACH, 0);
 587         kfree_skb(skb);
 588         read_unlock(&ipip6_lock);
 589 out:
 590         return 0;
 591 }
 592
 593 /* Returns the embedded IPv4 address if the IPv6 address
 594    comes from 6to4 (RFC 3056) addr space */
 595
 596 static inline __be32 try_6to4(struct in6_addr *v6dst)
 597 {
 598         __be32 dst = 0;
 599
 600         if (v6dst->s6_addr16[0] == htons(0x2002)) {
 601                 /* 6to4 v6 addr has 16 bits prefix, 32 v4addr, 16 SLA, ... */
 602                 memcpy(&dst, &v6dst->s6_addr16[1], 4);
 603         }
 604         return dst;
 605 }
 606
 607 /*
 608  *      This function assumes it is being called from dev_queue_xmit()
 609  *      and that skb is filled properly by that function.
 610  */
 611
 612 static int ipip6_tunnel_xmit(struct sk_buff *skb, struct net_device *dev)
 613 {
 614         struct ip_tunnel *tunnel = netdev_priv(dev);
 615         struct net_device_stats *stats = &tunnel->stat;
 616         struct iphdr  *tiph = &tunnel->parms.iph;
 617         struct ipv6hdr *iph6 = ipv6_hdr(skb);
 618         u8     tos = tunnel->parms.iph.tos;
 619         struct rtable *rt;                      /* Route to the other host */
 620         struct net_device *tdev;                        /* Device to other host */
 621         struct iphdr  *iph;                     /* Our new IP header */
 622         unsigned int max_headroom;              /* The extra header space needed */
 623         __be32 dst = tiph->daddr;
 624         int    mtu;
 625         struct in6_addr *addr6;
 626         int addr_type;
 627
 628         if (tunnel->recursion++) {
 629                 tunnel->stat.collisions++;
 630                 goto tx_error;
 631         }
 632
 633         if (skb->protocol != htons(ETH_P_IPV6))
 634                 goto tx_error;
 635
 636         /* ISATAP (RFC4214) - must come before 6to4 */
 637         if (dev->priv_flags & IFF_ISATAP) {
 638                 struct neighbour *neigh = NULL;
 639
 640                 if (skb->dst)
 641                         neigh = skb->dst->neighbour;
 642
 643                 if (neigh == NULL) {
 644                         if (net_ratelimit())
 645                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 646                         goto tx_error;
 647                 }
 648
 649                 addr6 = (struct in6_addr*)&neigh->primary_key;
 650                 addr_type = ipv6_addr_type(addr6);
 651
 652                 if ((addr_type & IPV6_ADDR_UNICAST) &&
 653                      ipv6_addr_is_isatap(addr6))
 654                         dst = addr6->s6_addr32[3];
 655                 else
 656                         goto tx_error;
 657         }
 658
 659         if (!dst)
 660                 dst = try_6to4(&iph6->daddr);
 661
 662         if (!dst) {
 663                 struct neighbour *neigh = NULL;
 664
 665                 if (skb->dst)
 666                         neigh = skb->dst->neighbour;
 667
 668                 if (neigh == NULL) {
 669                         if (net_ratelimit())
 670                                 printk(KERN_DEBUG "sit: nexthop == NULL\n");
 671                         goto tx_error;
 672                 }
 673
 674                 addr6 = (struct in6_addr*)&neigh->primary_key;
 675                 addr_type = ipv6_addr_type(addr6);
 676
 677                 if (addr_type == IPV6_ADDR_ANY) {
 678                         addr6 = &ipv6_hdr(skb)->daddr;
 679                         addr_type = ipv6_addr_type(addr6);
 680                 }
 681
 682                 if ((addr_type & IPV6_ADDR_COMPATv4) == 0)
 683                         goto tx_error_icmp;
 684
 685                 dst = addr6->s6_addr32[3];
 686         }
 687
 688         {
 689                 struct flowi fl = { .nl_u = { .ip4_u =
 690                                               { .daddr = dst,
 691                                                 .saddr = tiph->saddr,
 692                                                 .tos = RT_TOS(tos) } },
 693                                     .oif = tunnel->parms.link,
 694                                     .proto = IPPROTO_IPV6 };
 695                 if (ip_route_output_key(&init_net, &rt, &fl)) {
 696                         tunnel->stat.tx_carrier_errors++;
 697                         goto tx_error_icmp;
 698                 }
 699         }
 700         if (rt->rt_type != RTN_UNICAST) {
 701                 ip_rt_put(rt);
 702                 tunnel->stat.tx_carrier_errors++;
 703                 goto tx_error_icmp;
 704         }
 705         tdev = rt->u.dst.dev;
 706
 707         if (tdev == dev) {
 708                 ip_rt_put(rt);
 709                 tunnel->stat.collisions++;
 710                 goto tx_error;
 711         }
 712
 713         if (tiph->frag_off)
 714                 mtu = dst_mtu(&rt->u.dst) - sizeof(struct iphdr);
 715         else
 716                 mtu = skb->dst ? dst_mtu(skb->dst) : dev->mtu;
 717
 718         if (mtu < 68) {
 719                 tunnel->stat.collisions++;
 720                 ip_rt_put(rt);
 721                 goto tx_error;
 722         }
 723         if (mtu < IPV6_MIN_MTU)
 724                 mtu = IPV6_MIN_MTU;
 725         if (tunnel->parms.iph.daddr && skb->dst)
 726                 skb->dst->ops->update_pmtu(skb->dst, mtu);
 727
 728         if (skb->len > mtu) {
 729                 icmpv6_send(skb, ICMPV6_PKT_TOOBIG, 0, mtu, dev);
 730                 ip_rt_put(rt);
 731                 goto tx_error;
 732         }
 733
 734         if (tunnel->err_count > 0) {
 735                 if (jiffies - tunnel->err_time < IPTUNNEL_ERR_TIMEO) {
 736                         tunnel->err_count--;
 737                         dst_link_failure(skb);
 738                 } else
 739                         tunnel->err_count = 0;
 740         }
 741
 742         /*
 743          * Okay, now see if we can stuff it in the buffer as-is.
 744          */
 745         max_headroom = LL_RESERVED_SPACE(tdev)+sizeof(struct iphdr);
 746
 747         if (skb_headroom(skb) < max_headroom || skb_shared(skb) ||
 748             (skb_cloned(skb) && !skb_clone_writable(skb, 0))) {
 749                 struct sk_buff *new_skb = skb_realloc_headroom(skb, max_headroom);
 750                 if (!new_skb) {
 751                         ip_rt_put(rt);
 752                         stats->tx_dropped++;
 753                         dev_kfree_skb(skb);
 754                         tunnel->recursion--;
 755                         return 0;
 756                 }
 757                 if (skb->sk)
 758                         skb_set_owner_w(new_skb, skb->sk);
 759                 dev_kfree_skb(skb);
 760                 skb = new_skb;
 761                 iph6 = ipv6_hdr(skb);
 762         }
 763
 764         skb->transport_header = skb->network_header;
 765         skb_push(skb, sizeof(struct iphdr));
 766         skb_reset_network_header(skb);
 767         memset(&(IPCB(skb)->opt), 0, sizeof(IPCB(skb)->opt));
 768         IPCB(skb)->flags = 0;
 769         dst_release(skb->dst);
 770         skb->dst = &rt->u.dst;
 771
 772         /*
 773          *      Push down and install the IPIP header.
 774          */
 775
 776         iph                     =       ip_hdr(skb);
 777         iph->version            =       4;
 778         iph->ihl                =       sizeof(struct iphdr)>>2;
 779         if (mtu > IPV6_MIN_MTU)
 780                 iph->frag_off   =       htons(IP_DF);
 781         else
 782                 iph->frag_off   =       0;
 783
 784         iph->protocol           =       IPPROTO_IPV6;
 785         iph->tos                =       INET_ECN_encapsulate(tos, ipv6_get_dsfield(iph6));
 786         iph->daddr              =       rt->rt_dst;
 787         iph->saddr              =       rt->rt_src;
 788
 789         if ((iph->ttl = tiph->ttl) == 0)
 790                 iph->ttl        =       iph6->hop_limit;
 791
 792         nf_reset(skb);
 793
 794         IPTUNNEL_XMIT();
 795         tunnel->recursion--;
 796         return 0;
 797
 798 tx_error_icmp:
 799         dst_link_failure(skb);
 800 tx_error:
 801         stats->tx_errors++;
 802         dev_kfree_skb(skb);
 803         tunnel->recursion--;
 804         return 0;
 805 }
 806
 807 static void ipip6_tunnel_bind_dev(struct net_device *dev)
 808 {
 809         struct net_device *tdev = NULL;
 810         struct ip_tunnel *tunnel;
 811         struct iphdr *iph;
 812
 813         tunnel = netdev_priv(dev);
 814         iph = &tunnel->parms.iph;
 815
 816         if (iph->daddr) {
 817                 struct flowi fl = { .nl_u = { .ip4_u =
 818                                               { .daddr = iph->daddr,
 819                                                 .saddr = iph->saddr,
 820                                                 .tos = RT_TOS(iph->tos) } },
 821                                     .oif = tunnel->parms.link,
 822                                     .proto = IPPROTO_IPV6 };
 823                 struct rtable *rt;
 824                 if (!ip_route_output_key(&init_net, &rt, &fl)) {
 825                         tdev = rt->u.dst.dev;
 826                         ip_rt_put(rt);
 827                 }
 828                 dev->flags |= IFF_POINTOPOINT;
 829         }
 830
 831         if (!tdev && tunnel->parms.link)
 832                 tdev = __dev_get_by_index(&init_net, tunnel->parms.link);
 833
 834         if (tdev) {
 835                 dev->hard_header_len = tdev->hard_header_len + sizeof(struct iphdr);
 836                 dev->mtu = tdev->mtu - sizeof(struct iphdr);
 837                 if (dev->mtu < IPV6_MIN_MTU)
 838                         dev->mtu = IPV6_MIN_MTU;
 839         }
 840         dev->iflink = tunnel->parms.link;
 841 }
 842
 843 static int
 844 ipip6_tunnel_ioctl (struct net_device *dev, struct ifreq *ifr, int cmd)
 845 {
 846         int err = 0;
 847         struct ip_tunnel_parm p;
 848         struct ip_tunnel_prl prl;
 849         struct ip_tunnel *t;
 850
 851         switch (cmd) {
 852         case SIOCGETTUNNEL:
 853                 t = NULL;
 854                 if (dev == ipip6_fb_tunnel_dev) {
 855                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p))) {
 856                                 err = -EFAULT;
 857                                 break;
 858                         }
 859                         t = ipip6_tunnel_locate(&p, 0);
 860                 }
 861                 if (t == NULL)
 862                         t = netdev_priv(dev);
 863                 memcpy(&p, &t->parms, sizeof(p));
 864                 if (copy_to_user(ifr->ifr_ifru.ifru_data, &p, sizeof(p)))
 865                         err = -EFAULT;
 866                 break;
 867
 868         case SIOCADDTUNNEL:
 869         case SIOCCHGTUNNEL:
 870                 err = -EPERM;
 871                 if (!capable(CAP_NET_ADMIN))
 872                         goto done;
 873
 874                 err = -EFAULT;
 875                 if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 876                         goto done;
 877
 878                 err = -EINVAL;
 879                 if (p.iph.version != 4 || p.iph.protocol != IPPROTO_IPV6 ||
 880                     p.iph.ihl != 5 || (p.iph.frag_off&htons(~IP_DF)))
 881                         goto done;
 882                 if (p.iph.ttl)
 883                         p.iph.frag_off |= htons(IP_DF);
 884
 885                 t = ipip6_tunnel_locate(&p, cmd == SIOCADDTUNNEL);
 886
 887                 if (dev != ipip6_fb_tunnel_dev && cmd == SIOCCHGTUNNEL) {
 888                         if (t != NULL) {
 889                                 if (t->dev != dev) {
 890                                         err = -EEXIST;
 891                                         break;
 892                                 }
 893                         } else {
 894                                 if (((dev->flags&IFF_POINTOPOINT) && !p.iph.daddr) ||
 895                                     (!(dev->flags&IFF_POINTOPOINT) && p.iph.daddr)) {
 896                                         err = -EINVAL;
 897                                         break;
 898                                 }
 899                                 t = netdev_priv(dev);
 900                                 ipip6_tunnel_unlink(t);
 901                                 t->parms.iph.saddr = p.iph.saddr;
 902                                 t->parms.iph.daddr = p.iph.daddr;
 903                                 memcpy(dev->dev_addr, &p.iph.saddr, 4);
 904                                 memcpy(dev->broadcast, &p.iph.daddr, 4);
 905                                 ipip6_tunnel_link(t);
 906                                 netdev_state_change(dev);
 907                         }
 908                 }
 909
 910                 if (t) {
 911                         err = 0;
 912                         if (cmd == SIOCCHGTUNNEL) {
 913                                 t->parms.iph.ttl = p.iph.ttl;
 914                                 t->parms.iph.tos = p.iph.tos;
 915                                 if (t->parms.link != p.link) {
 916                                         t->parms.link = p.link;
 917                                         ipip6_tunnel_bind_dev(dev);
 918                                         netdev_state_change(dev);
 919                                 }
 920                         }
 921                         if (copy_to_user(ifr->ifr_ifru.ifru_data, &t->parms, sizeof(p)))
 922                                 err = -EFAULT;
 923                 } else
 924                         err = (cmd == SIOCADDTUNNEL ? -ENOBUFS : -ENOENT);
 925                 break;
 926
 927         case SIOCDELTUNNEL:
 928                 err = -EPERM;
 929                 if (!capable(CAP_NET_ADMIN))
 930                         goto done;
 931
 932                 if (dev == ipip6_fb_tunnel_dev) {
 933                         err = -EFAULT;
 934                         if (copy_from_user(&p, ifr->ifr_ifru.ifru_data, sizeof(p)))
 935                                 goto done;
 936                         err = -ENOENT;
 937                         if ((t = ipip6_tunnel_locate(&p, 0)) == NULL)
 938                                 goto done;
 939                         err = -EPERM;
 940                         if (t == netdev_priv(ipip6_fb_tunnel_dev))
 941                                 goto done;
 942                         dev = t->dev;
 943                 }
 944                 unregister_netdevice(dev);
 945                 err = 0;
 946                 break;
 947
 948         case SIOCGETPRL:
 949         case SIOCADDPRL:
 950         case SIOCDELPRL:
 951         case SIOCCHGPRL:
 952                 err = -EPERM;
 953                 if (cmd != SIOCGETPRL && !capable(CAP_NET_ADMIN))
 954                         goto done;
 955                 err = -EINVAL;
 956                 if (dev == ipip6_fb_tunnel_dev)
 957                         goto done;
 958                 err = -EFAULT;
 959                 if (copy_from_user(&prl, ifr->ifr_ifru.ifru_data, sizeof(prl)))
 960                         goto done;
 961                 err = -ENOENT;
 962                 if (!(t = netdev_priv(dev)))
 963                         goto done;
 964
 965                 switch (cmd) {
 966                 case SIOCGETPRL:
 967                         err = ipip6_tunnel_get_prl(t, &prl);
 968                         if (!err && copy_to_user(ifr->ifr_ifru.ifru_data,
 969                                                  &prl, sizeof(prl)))
 970                                 err = -EFAULT;
 971                         break;
 972                 case SIOCDELPRL:
 973                         err = ipip6_tunnel_del_prl(t, &prl);
 974                         break;
 975                 case SIOCADDPRL:
 976                 case SIOCCHGPRL:
 977                         err = ipip6_tunnel_add_prl(t, &prl, cmd == SIOCCHGPRL);
 978                         break;
 979                 }
 980                 if (cmd != SIOCGETPRL)
 981                         netdev_state_change(dev);
 982                 break;
 983
 984         default:
 985                 err = -EINVAL;
 986         }
 987
 988 done:
 989         return err;
 990 }
 991
 992 static struct net_device_stats *ipip6_tunnel_get_stats(struct net_device *dev)
 993 {
 994         return &(((struct ip_tunnel*)netdev_priv(dev))->stat);
 995 }
 996
 997 static int ipip6_tunnel_change_mtu(struct net_device *dev, int new_mtu)
 998 {
 999         if (new_mtu < IPV6_MIN_MTU || new_mtu > 0xFFF8 - sizeof(struct iphdr))
1000                 return -EINVAL;
1001         dev->mtu = new_mtu;
1002         return 0;
1003 }
1004
1005 static void ipip6_tunnel_setup(struct net_device *dev)
1006 {
1007         dev->uninit             = ipip6_tunnel_uninit;
1008         dev->destructor         = free_netdev;
1009         dev->hard_start_xmit    = ipip6_tunnel_xmit;
1010         dev->get_stats          = ipip6_tunnel_get_stats;
1011         dev->do_ioctl           = ipip6_tunnel_ioctl;
1012         dev->change_mtu         = ipip6_tunnel_change_mtu;
1013
1014         dev->type               = ARPHRD_SIT;
1015         dev->hard_header_len    = LL_MAX_HEADER + sizeof(struct iphdr);
1016         dev->mtu                = ETH_DATA_LEN - sizeof(struct iphdr);
1017         dev->flags              = IFF_NOARP;
1018         dev->iflink             = 0;
1019         dev->addr_len           = 4;
1020 }
1021
1022 static int ipip6_tunnel_init(struct net_device *dev)
1023 {
1024         struct ip_tunnel *tunnel;
1025
1026         tunnel = netdev_priv(dev);
1027
1028         tunnel->dev = dev;
1029         strcpy(tunnel->parms.name, dev->name);
1030
1031         memcpy(dev->dev_addr, &tunnel->parms.iph.saddr, 4);
1032         memcpy(dev->broadcast, &tunnel->parms.iph.daddr, 4);
1033
1034         ipip6_tunnel_bind_dev(dev);
1035
1036         return 0;
1037 }
1038
1039 static int __init ipip6_fb_tunnel_init(struct net_device *dev)
1040 {
1041         struct ip_tunnel *tunnel = netdev_priv(dev);
1042         struct iphdr *iph = &tunnel->parms.iph;
1043
1044         tunnel->dev = dev;
1045         strcpy(tunnel->parms.name, dev->name);
1046
1047         iph->version            = 4;
1048         iph->protocol           = IPPROTO_IPV6;
1049         iph->ihl                = 5;
1050         iph->ttl                = 64;
1051
1052         dev_hold(dev);
1053         tunnels_wc[0]           = tunnel;
1054         return 0;
1055 }
1056
1057 static struct xfrm_tunnel sit_handler = {
1058         .handler        =       ipip6_rcv,
1059         .err_handler    =       ipip6_err,
1060         .priority       =       1,
1061 };
1062
1063 static void __exit sit_destroy_tunnels(void)
1064 {
1065         int prio;
1066
1067         for (prio = 1; prio < 4; prio++) {
1068                 int h;
1069                 for (h = 0; h < HASH_SIZE; h++) {
1070                         struct ip_tunnel *t;
1071                         while ((t = tunnels[prio][h]) != NULL)
1072                                 unregister_netdevice(t->dev);
1073                 }
1074         }
1075 }
1076
1077 static int sit_init_net(struct net *net)
1078 {
1079         int err;
1080         struct sit_net *sitn;
1081
1082         err = -ENOMEM;
1083         sitn = kmalloc(sizeof(struct sit_net), GFP_KERNEL);
1084         if (sitn == NULL)
1085                 goto err_alloc;
1086
1087         err = net_assign_generic(net, sit_net_id, sitn);
1088         if (err < 0)
1089                 goto err_assign;
1090
1091         return 0;
1092
1093 err_assign:
1094         kfree(sitn);
1095 err_alloc:
1096         return err;
1097 }
1098
1099 static void sit_exit_net(struct net *net)
1100 {
1101         struct sit_net *sitn;
1102
1103         sitn = net_generic(net, sit_net_id);
1104         kfree(sitn);
1105 }
1106
1107 static struct pernet_operations sit_net_ops = {
1108         .init = sit_init_net,
1109         .exit = sit_exit_net,
1110 };
1111
1112 static void __exit sit_cleanup(void)
1113 {
1114         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1115
1116         rtnl_lock();
1117         sit_destroy_tunnels();
1118         unregister_netdevice(ipip6_fb_tunnel_dev);
1119         rtnl_unlock();
1120
1121         unregister_pernet_gen_device(sit_net_id, &sit_net_ops);
1122 }
1123
1124 static int __init sit_init(void)
1125 {
1126         int err;
1127
1128         printk(KERN_INFO "IPv6 over IPv4 tunneling driver\n");
1129
1130         if (xfrm4_tunnel_register(&sit_handler, AF_INET6) < 0) {
1131                 printk(KERN_INFO "sit init: Can't add protocol\n");
1132                 return -EAGAIN;
1133         }
1134
1135         ipip6_fb_tunnel_dev = alloc_netdev(sizeof(struct ip_tunnel), "sit0",
1136                                            ipip6_tunnel_setup);
1137         if (!ipip6_fb_tunnel_dev) {
1138                 err = -ENOMEM;
1139                 goto err1;
1140         }
1141
1142         ipip6_fb_tunnel_dev->init = ipip6_fb_tunnel_init;
1143
1144         if ((err =  register_netdev(ipip6_fb_tunnel_dev)))
1145                 goto err2;
1146
1147         err = register_pernet_gen_device(&sit_net_id, &sit_net_ops);
1148         if (err < 0)
1149                 goto err3;
1150
1151  out:
1152         return err;
1153  err2:
1154         free_netdev(ipip6_fb_tunnel_dev);
1155  err1:
1156         xfrm4_tunnel_deregister(&sit_handler, AF_INET6);
1157         goto out;
1158 err3:
1159         unregister_netdevice(ipip6_fb_tunnel_dev);
1160         goto err1;
1161 }
1162
1163 module_init(sit_init);
1164 module_exit(sit_cleanup);
1165 MODULE_LICENSE("GPL");
1166 MODULE_ALIAS("sit0");